Utility function to assess conditional independence #791
Conversation
There was a problem hiding this comment.
This is excellent work. Unit test, inits, and function are well-written. Only minor suggestions requested on formatting.
If I understand correctly this is Alg 1 from Shachter (1998), which is somewhat inefficient. It's a great starting point though for a future PR.
edward/util/random_variables.py
Outdated
| bool | ||
| True if a is independent of b given the random variables in condition. | ||
|
|
||
| References |
There was a problem hiding this comment.
Can you add a bibtex entry to docs/tex/bib.bib? Note the bib file is organized chronologically by year, then alphabetically by author's last name within a year. For how to cite it in the docstring, see e.g., edward/criticisms/ppc.py.
edward/util/random_variables.py
Outdated
|
|
||
| Implemented using the Bayes-Ball algorithm[1]. | ||
|
|
||
| Args |
There was a problem hiding this comment.
Fields like Args and Returns require a colon at the end, c.f., other function docstrings in this file.
edward/util/random_variables.py
Outdated
| Implemented using the Bayes-Ball algorithm[1]. | ||
|
|
||
| Args | ||
| a : RandomVariable or list of RandomVariable |
There was a problem hiding this comment.
Args are written as, e.g., a: RandomVariable or list of RandomVariable. Note the removed space and added period. When in doubt, you can also double check the generated API docstrings by checking out the README in docs/, or consult with the other function docstrings.
|
Thank you very much for the feedback. I've made changes to the docstring and added a bibtex entry. The function actually uses Alg 2 from Schachter (1998). If you can point me towards a more efficient approach I'd be happy to modify the implementation. |
|
Perfect! LGTM. |
This is an attempt at addressing issue #290. The utility function
is_independentis located atedward/util/random_variables.pyand I've added some tests attests/util/test_is_independent.